MCP Servers
Connect Model Context Protocol servers to BullSequana AI and reach them through the MCP Gateway.
An MCP Server is a tool endpoint that BullSequana AI makes available to one organization. A Platform Admin adds the server and chooses which of its tools may be called. Each member then links their own credential to it, and only after that can they see or invoke its tools.
Every call to an MCP Server passes through the MCP Gateway. Nothing reaches an MCP Server another way.
In this section
Understand the model
Administer
Add an MCP Server
Register an external endpoint, or run a container inside the platform.
Restrict the tools
Choose which of a server's tools the organization may call.
Use
Link your credential
Sign in to the upstream system with OAuth, or paste a token.
Use an MCP Server in chat
Turn a connector on for a conversation and watch the model call its tools.
Connect an MCP client
Point a coding agent or a desktop MCP client at the Gateway.
Reference
Who does what
| Task | Platform Admin | Member |
|---|---|---|
| Add, edit, and delete an MCP Server | Yes | No |
| Set the tool allowlist | Yes | No |
| Read the tools a server advertises | Yes | No |
| See the servers the organization has | Yes | Yes |
| Link and unlink their own credential | Yes | Yes |
| Call a server through the Gateway | Yes | Yes |
A Platform Admin holds the admin role in the organization. Members hold any role that grants API access.
Where MCP Servers are used
An MCP Server reaches BullSequana AI users in two places.
An Agent can list the MCP Servers it may reach. When a member runs that Agent in chat, the tools of every server they have linked become available to the model. Servers they have not linked are absent from the run.
An external MCP client, such as a coding agent or a desktop assistant, can connect straight to a server's Gateway URL. See Connect an MCP client.